Observing field model. Not drawn to scale.

Published in 2002 by the CIE Technical Committee 8-01 (Color Appearance Modelling for Color Management Systems), as of 2008 CIECAM02 is the most recent color appearance model ratified by the CIE, and the successor of CIECAM97s.^[1]

The two major parts of the model are its chromatic adaptation transform, CIECAT02, and its equations for calculating mathematical correlates for the six technically-defined dimensions of color appearance: brightness (luminance), lightness, colorfulness, chroma, saturation, and hue.

CIECAM02 takes for its input the tristimulus values of the stimulus, the tristimulus values of an adapting white point, adapting background, and surround luminance information, and whether or not observers are discounting the illuminant (color constancy is in effect). The model can be used to predict these appearance attributes or, with forward and reverse implementations for distinct viewing conditions, to compute corresponding colors.

CIECAM02 is used in Windows Vista's Windows Color System.^[2]

Viewing conditions[link]

The inner circle is the stimulus, from which the tristimulus values should be measured in CIE XYZ using the 2° standard observer. The intermediate circle is the proximal field, extending out another 2°. The outer circle is the background, reaching out to 10°, from which the relative luminance (Y_b) need be measured. If the proximal field is the same color as the background, the background is considered to be adjacent to the stimulus. Beyond the circles which comprise the display field (display area, viewing area) is the surround field (or peripheral area), which can be considered to be the entire room. The totality of the proximal field, background, and surround is called the adapting field (the field of view that supports adaptation—extends to the limit of vision).^[3]

When referring to the literature, it is also useful to be aware of the difference between the terms adopted white point (the computational white point) and the adapted white point (the observer white point).^[4] The distinction may be important in mixed mode illumination, where psychophysical phenomena come into play. This is a subject of research.

Parameter decision table[link]

CIECAM02 defines three surround(ing)s – average, dim, and dark – with associated parameters defined here for reference in the rest of this article:^[5]

• S_R = L_sw / L_dw: ratio of the absolute luminance of the reference white (white point) measured in the surround field to the display area. The 0.2 coefficient derives from the "gray world" assumption (~18%–20% reflectivity). It tests whether the surround luminance is darker or brighter than medium gray.
• F: factor determining degree of adaptation
• c: impact of surrounding
• N_c: chromatic induction factor

For intermediate conditions, these values can be linearly interpolated.^[5]

The absolute luminance of the adapting field, which is a quantity that will be needed later, should be measured with a photometer. If one is not available, it can be calculated using a reference white:

Failed to parse (Missing texvc executable; please see math/README to configure.): L_A = \frac{E_w}{\pi} \frac{Y_b}{Y_w} = \frac{L_W Y_b}{Y_w}

where Y_b is the relative luminance of background, the E_w = πL_W is the illuminance of the reference white in lux, L_W is the absolute luminance of the reference white in cd/m², and Y_w is the relative luminance of the reference white in the adapting field. If unknown, the adapting field can be assumed to have average reflectance ("gray world" assumption): L_A = L_W / 5.

Note: Care should be taken not to confuse L_W, the absolute luminance of the reference white in cd/m², and L_w the red cone response in the LMS color space.

Chromatic adaptation[link]

Summary[link]

1. Convert to the "spectrally sharpened" CAT02 LMS space to prepare for adaptation. Spectral sharpening is the transformation of the tristimulus values into new values that would have resulted from a sharper, more concentrated set of spectral sensitivities. It is argued that this aids color constancy, especially in the blue region.^{[citation needed]}
2. Perform chromatic adaptation using CAT02 (also known as the "modified CMCCAT2000 transform").
3. Convert to an LMS space closer to the cone fundamentals. It is argued that predicting perceptual attribute correlates is best done in such spaces.^[5]
4. Perform post-adaptation cone response compression.

CAT02[link]

Given a set of tristimulus values in XYZ, the corresponding LMS values can be determined by the M_CAT02 transformation matrix (calculated using the CIE 1931 2° standard colorimetric observer).^[1] The sample color in the test illuminant is:

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{bmatrix} L\\ M\\ S \end{bmatrix} = \mathbf{M}_{CAT02} \begin{bmatrix} X\\ Y\\ Z \end{bmatrix},\quad \mathbf{M}_{CAT02} = \begin{bmatrix} \;\;\,0.7328 & 0.4296 & -0.1624\\ -0.7036 & 1.6975 & \;\;\,0.0061\\ \;\;\,0.0030 & 0.0136 & \;\;\,0.9834 \end{bmatrix}

Once in LMS, the white point can be adapted to the desired degree by choosing the parameter D.^[3] For the general CAT02, the corresponding color in the reference illuminant is:

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{align} L_c &= \Big(\frac{Y_w L_{wr}}{Y_{wr} L_w} D + 1-D\Big)L\\ M_c &=\Big(\frac{Y_w M_{wr}}{Y_{wr} M_w} D + 1-D\Big)M\\ S_c &= \Big(\frac{Y_w S_{wr}}{Y_{wr} S_w} D + 1-D\Big)S\\ \end{align}

where the Y_w / Y_wr factor accounts for the two illuminants having the same chromaticity but different reference whites.^[6] The subscripts indicate the cone response for white under the test (w) and reference illuminant (wr). The degree of adaptation (discounting) D can be set to zero for no adaptation (stimulus is considered self-luminous) and unity for complete adaptation (color constancy). In practice, it ranges from 0.65 to 1.0, as can be seen from the diagram. Intermediate values can be calculated by:^[5]

Failed to parse (Missing texvc executable; please see math/README to configure.): D = F \left( 1 - \textstyle{\frac{1}{3.6}} e^{-(L_A + 42) / 92} \right)

where surround F is as defined above and L_A is the adapting field luminance in cd/m².^[1]

log-log plot of F_L vs. L_A (L_A ranges from 10⁻⁴ to 10⁴, F_L ranges from 10⁻⁴ to 10). The cube root approximation of F_L is 0.1715L_A^1/3

In CIECAM02, the reference illuminant has equal energy L_wr = M_wr = S_wr = 100) and the reference white is the perfect reflecting diffuser (i.e., unity reflectance, and Y_wr = 100) hence:

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{align} L_c &= \Big(\frac{Y_w}{L_w} D + 1-D\Big)L\\ M_c &=\Big(\frac{Y_w}{M_w} D + 1-D\Big)M\\ S_c &= \Big(\frac{Y_w}{S_w} D + 1-D\Big)S\\ \end{align}

Furthermore, if the reference white in both illuminants have the Y tristimulus value (Y_wr = Y_w) then:

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{align} L_c &= \Big(\frac{L_{wr}}{L_w} D + 1-D\Big)L\\ M_c &=\Big(\frac{M_{wr}}{M_w} D + 1-D\Big)M\\ S_c &= \Big(\frac{S_{wr}}{S_w} D + 1-D\Big)S\\ \end{align}

Post-adaptation[link]

After adaptation, the cone responses are converted to the Hunt–Pointer–Estévez space by going to XYZ and back:^[5]

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{bmatrix} L' \\ M' \\ S' \end{bmatrix} = \mathbf{M}_H \begin{bmatrix} X_c \\ Y_c \\ Z_c \end{bmatrix} = \mathbf{M}_H \mathbf{M}_{CAT02}^{-1} \begin{bmatrix} L_c \\ M_c \\ S_c \end{bmatrix}

Failed to parse (Missing texvc executable; please see math/README to configure.): \mathbf{M}_H = \begin{bmatrix} \;\;\,0.38971 & 0.68898 & -0.07868 \\ -0.22981 & 1.18340 & \;\;\,0.04641 \\ \;\;\,0.00000 & 0.00000 & \;\;\,1.00000 \end{bmatrix}

log L′_a vs. log L′ for L_A = 200 (F_L = 1)

Finally, the response is compressed based on the generalized Michaelis–Menten equation (as depicted aside):^[5]

Failed to parse (Missing texvc executable; please see math/README to configure.): k = \frac{1}{5 L_A + 1}

Failed to parse (Missing texvc executable; please see math/README to configure.): F_L = \textstyle{\frac{1}{5}} k^4 \left( 5 L_A \right) + \textstyle{\frac{1}{10}} {(1 - k^4)}^2 {\left( 5 L_A \right)}^{1/3}

F_L is the luminance level adaptation factor.

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{align} L'_a &= \frac{400 {\left(F_L L'/100\right)}^{0.42}}{27.13 + {\left(F_L L'/100\right)}^{0.42}} + 0.1 \\ M'_a &= \frac{400 {\left(F_L M'/100\right)}^{0.42}}{27.13 + {\left(F_L M'/100\right)}^{0.42}} + 0.1 \\ S'_a &= \frac{400 {\left(F_L S'/100\right)}^{0.42}}{27.13 + {\left(F_L S'/100\right)}^{0.42}} + 0.1 \end{align}

As previously mentioned, if the luminance level of the background is unknown, it can estimated from the absolute luminance of the white point as L_A = L_W / 5 using the "medium gray" assumption. (The expression for F_L is given in terms of 5L_A for convenience.) In photopic conditions, the luminance level adaptation factor (F_L) is proportional to the cube root of the luminance of the adapting field (L_A). In scotopic conditions, it is proportional to L_A (meaning no luminance level adaptation). The photopic threshold is roughly L_W = 1 (see F_L–L_A graph above).

Appearance correlates[link]

CIECAM02 defines correlates for yellow-blue, red-green, brightness, and colorfulness. Let us make some preliminary definitions.

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{align} C_1 &= L^\prime_a - M^\prime_a \\ C_2 &= M^\prime_a - S^\prime_a \\ C_3 &= S^\prime_a - L^\prime_a \end{align}

The correlate for red–green (a) is the magnitude of the departure of C₁ from the criterion for unique yellow (C₁ = C₂ / 11), and the correlate for yellow–blue (b) is based on the mean of the magnitude of the departures of C₁ from unique red (C₁ = C₂) and unique green (C₁ = C₃).^[3]

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{align} a &= C_1 - \textstyle{\frac{1}{11}}C_2 &= L^\prime_a - \textstyle{\frac{12}{11}} M^\prime_a + \textstyle{\frac{1}{11}} S^\prime_a \\ b &= \textstyle{\frac{1}{2}} \left( C_2 - C_1 + C_1 - C_3 \right) / 4.5 &= \textstyle{\frac{1}{9}} \left( L^\prime_a + M^\prime_a - 2S^\prime_a \right) \end{align}

The 4.5 factor accounts for the fact that there are fewer cones at shorter wavelengths (the eye is less sensitive to blue). The order of the terms is such that b is positive for yellowish colors (rather than blueish).

The hue angle (h) can be found by converting the rectangular coordinate (a, b) into polar coordinates:

Failed to parse (Missing texvc executable; please see math/README to configure.): h = \angle (a, b),\ (0 < h < 360^\circ)

To calculate the eccentricity (e_t) and hue composition (H), determine which quadrant the hue is in with the aid of the following table. Choose i such that h_i ≤ h′ < h_i+1, where h′ = h if h > h₁ and h′ = h + 360° otherwise.

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{align} H &= H_i + \frac{100 (h^\prime - h_i) / e_i}{(h^\prime - h_i) / e_i + (h_{i+1} - h^\prime) / e_{i+1}} \\ e_t &= \textstyle{\frac{1}{4}} \left[ \cos\left( \textstyle{\frac{\pi}{180}}h + 2\right) + 3.8 \right] \end{align}

(This is not exactly the same as the eccentricity factor given in the table.)

Calculate the achromatic response A:

Failed to parse (Missing texvc executable; please see math/README to configure.): A = (2 L^\prime_a + M^\prime_a + \textstyle{\frac{1}{20}} S^\prime_a - 0.305) N_{bb}

where

Failed to parse (Missing texvc executable; please see math/README to configure.): \begin{align} &N_{bb} = N_{cb} = 0.725 n^{-0.2} \\ &n = Y_b / Y_w \end{align}

The correlate of lightness is

Failed to parse (Missing texvc executable; please see math/README to configure.): J = 100 \left( A / A_w \right)^{c z}

where c is the impact of surround (see above), and

Failed to parse (Missing texvc executable; please see math/README to configure.): z = 1.48 + \sqrt{n}

The correlate of brightness is

Failed to parse (Missing texvc executable; please see math/README to configure.): Q = \left(4 / c \right) \sqrt{\textstyle{\frac{1}{100}} J} \left(A_w + 4\right) F_L^{1/4}

Then calculate a temporary quantity t,

Failed to parse (Missing texvc executable; please see math/README to configure.): t = \frac{ \textstyle{\frac{50\,000}{13}} N_c N_{cb} e_t \sqrt{a^2+b^2} } { L_a^\prime + M_a^\prime + \textstyle{\frac{21}{20}} S_a^\prime }

The correlate of chroma is

Failed to parse (Missing texvc executable; please see math/README to configure.): C = t^{0.9} \sqrt {\textstyle{\frac{1}{100}} J} (1.64 - 0.29^n)^{0.73}

The correlate of colorfulness is

Failed to parse (Missing texvc executable; please see math/README to configure.): M = C \cdot F_L^{1/4}

The correlate of saturation is

Failed to parse (Missing texvc executable; please see math/README to configure.): s = 100 \sqrt {M / Q}

References[link]

1. ^ ^{a b c} Fairchild, Mark D.; Luo, M. R.; Hunt, R. W. G. (August 2000). "A Revision of CIECAM97s for Practical Applications". Color Research & Applications (Wiley Interscience) 25 (4): 260–266. DOI:10.1002/1520-6378(200008)25:4<260::AID-COL6>3.0.CO;2-9. http://www.cis.rit.edu/fairchild/PDFs/PAP10.pdf. | quote=The CIECAM97s model was adopted by the CIE in 1997 for color imaging applications. It includes forward and reverse modes. Some problems in using this model were found in recent field trials. This article suggests revision to the model in two respects: (a) to make the lightness (J) zero when the Y tristimulus value is zero, under all surround conditions; (b) to modify the chromatic induction factor (Nc) from 1.10 to 0.95 for the dim surround condition. To avoid confusion, it is suggested that the revised version of the model be designated CAM97s2. The article also describes an alternative mode to achieve a more nearly exact reversibility between the forward and reverse modes.
2. ^ “Windows Color System: The Next Generation Color Management System”. Microsoft white paper. September 13, 2005.
3. ^ ^{a b c} Schanda, János (2007). "The Future of Colorimetry in the CIE: Color Appearance". Colorimetry: Understanding the CIE System. Wiley Interscience. p. 359. ISBN 978-0-470-04904-4. 
4. ^ Westland, Stephen; Ripamonti, Caterina (2004). Computational Colour Science Using MATLAB. John Wiley & Sons. ISBN 0-470-84562-7. http://books.google.com/?id=zgeuoz_muCIC&pg=PA86&lpg=PA86&dq=%22adapting+field%22+background+environment. 
5. ^ ^{a b c d e f} Moroney, Nathan; Fairchild, Mark D.; Hunt, Robert W.G.; Li, Changjun; Luo, M. Ronnier; Newman, Todd (November 12 2002). "The CIECAM02 Color Appearance Model". IS&T/SID Tenth Color Imaging Conference. Scottsdale, Arizona: The Society for Imaging Science and Technology. ISBN 0-89208-241-0. http://www.polybytes.com/misc/Meet_CIECAM02.pdf. 
6. ^ Hunt, Robert W. G.; Changjun Li, M. Ronnier Luo (February 2005). "Chromatic Adaptation Transforms". Color Research & Applications (Wiley Interscience) 30 (1): 69. DOI:10.1002/col.20085. "Chromatic adaptation transforms (CATs) have appeared in different forms. The reasons for these forms, and the relationships between them, are described. The factors governing which type of CAT should be used in different applications are explained" 

• CIE TC 8-01 (2004). A Color appearance model for color management systems. Publication 159. Vienna: CIE Central Bureau. ISBN 3-901906-29-0. http://www.cie.co.at/publ/abst/159-04.html. 
• Fairchild, Mark D. (2004-11-09). "Color Appearance Models: CIECAM02 and Beyond". IS&T/SID 12th Color Imaging Conference. http://www.cis.rit.edu/fairchild/PDFs/AppearanceLec.pdf. Retrieved 2008-02-11 

External links[link]

• Excel spreadsheet with forward and inverse examples, by Eric Walowit and Grit O'Brien
• Experimental Implementation of the CIECAM02 Color Appearance Model in a Photoshop Compatible Plug-in (Windows Only), by Cliff Rames.
• Notes on the CIECAM02 Colour Appearance Model. Source code in C of the forward and reverse transforms, by Billy Biggs.
• CIECAM02 Java applet, by Nathan Moroney